Verdent Manager Coordinates Idea-to-Deployed App Builds
AICodeKinggo watch the original →
the gist
Verdent Manager decomposes high-level prompts into parallel tasks, applies long-term memory for preferences, selects specialist skills, enables Slack/Telegram control, and adds Eco Mode plus BYOK.
Manager Workflow
Verdent Manager sits above the standard agent workflow and coordinates full projects. Users provide a high-level goal, such as "build a waitlist app for my product idea with a landing page, an email capture form, an admin view, and basic deployment." Manager decomposes the goal into phases like requirements, UI structure, recommendation logic, email capture, validation, and testing. It dispatches parallel workers for each piece, tracks progress on a board with statuses for in-progress, waiting for review, and done, then presents results for approval.
Memory, Skills, and Preferences
Manager maintains long-term memory of user stack (for example, TypeScript, Tailwind, Supabase, Vercel), naming style, product details, pet peeves, and prior feedback. It applies these automatically across projects. Skills provide reusable workflows for testing, deployment, front-end design, security review, and GitHub actions; Manager selects and invokes the appropriate skill based on context, such as QA checklists or security reviews.
Remote Access and Personalization
Manager integrates with Slack and Telegram, allowing users to assign tasks remotely via messages like "deploy the latest version to staging and report back" or "take a look at the latest PR and tell me if anything risky changed." Users customize a nickname and avatar to make Manager resemble a team member.
Cost Management
Eco Mode employs lower-cost models for iterative, exploratory, or lightweight tasks like polishing copy or small changes. BYOK lets users add API keys from Anthropic, OpenAI, or OpenRouter, making those models available in chat selectors, presets, sub-agents, and reviewers.